About me
I run the test desk at GirlfriendsAI and score every app on the site against the same 50-point rubric. I pay for the plans I review, keep a fresh account per app, and run the same scripted conversations so the numbers can be compared from one review to the next.
I have been using AI chat products since the early GPT-3 demos, and the part that still interests me is how fast they change. A model swap or a quiet price bump can move a score by a full point, so when an app changes I re-test it and update the page. On the blog I check every price, plan and score against our reviews before a post goes live.
How I test
- One paid account per app, opened with a fresh email, on the plan the review names.
- A scripted 200-message conversation, with planted facts to measure memory at day ten.
- An explicit pass that counts refusals and moments where the system prompt leaks into the reply.
- Image and voice tests where the app offers them, plus a read of the privacy policy.
The full rubric is on the methodology page. To flag an error in one of my posts, write to hello@girlfriendsai.net with the URL and the line.